Expert system-based post-stroke robotic rehabilitation for hemiparetic arm

Abstract

Approximately 50 to 60 percent of the more than five million stroke survivors are moderately or minimally impaired, and may greatly benefit from rehabilitation. It is widely accepted that most of the motor recovery occurs within the first one year from the onset of stroke and tends to plateau soon after. However, in recent years, clinical studies have shown that chronic stroke patients can have motor recovery even after four to ten years. Hence, there is a strong need for cost-effective, long-term rehabilitation solutions. Long-term rehabilitation requires the therapist to provide repetitive movements to the affected limb. In order to provide the repetitive movements, robotic devices have been developed. However, the few commercially available robots lack comprehensive rehabilitation software. Therapists are required to spend a considerable amount of time programming the robot, monitoring the patients, analyzing the data from the robot, and assessing the progress of the patients. This dissertation focuses on designing, developing, and clinically testing an expert system-based post-stroke robotic rehabilitation system for hemiparetic arm. The expert system and the associated software tools help in testing the patient, suggesting treatment options, training the patient, analyzing the data from the robot, and monitoring the patient’s progress.

Perhaps the most important step in developing a rehabilitation system is to understand the clinical practices of stroke treatment. In order to accomplish this, interviews and discussions were conducted with physical and occupational therapists. Based on their input, a survey was conducted among the physical and occupational therapists in Kansas and Missouri. The survey respondents answered many questions regarding stroke rehabilitation in general, as well as robotic rehabilitation of the upper limb. After analyzing the survey responses and perusing the current literature, a robotic rehabilitation treatment protocol was developed with the help of therapists. This treatment protocol was implemented as a rule-based, forward chaining expert system using CLIPS. The associated tools, such as the training and testing programs, were developed using Tcl/Tk, and the data analysis program was developed using C programming language. The expert system-based robotic rehabilitation system underwent a clinical pilot study with two stroke patients.

The clinical study showed that the expert system could produce valuable suggestions to the therapist regarding the treatment options. The developed data analysis tool made it possible for the therapist to administer therapy with minimal supervision by producing a quick summary at the end of each training session. Based on the feedback from the patients it was evident that the robotic training programs were entertaining. The study also showed that robotic rehabilitation is beneficial even for chronic stroke patients. The results of this research clearly suggest that it is not necessary for a therapist to continuously monitor a stroke patient during robotic training. Given the proper software tools for a rehabilitation robot, cost-effective long-term therapy can be delivered with minimal supervision.